University of Groningen
Effect of F0 contours on top-down repair of interrupted speech
Clarke, Jeanne; Kazanoglu, Deniz; Baskent, Deniz; Gaudrain, Etienne
Published in:
Journal of the Acoustical Society of America DOI:
10.1121/1.4990398
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2017
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Clarke, J., Kazanoglu, D., Baskent, D., & Gaudrain, E. (2017). Effect of F0 contours on top-down repair of interrupted speech. Journal of the Acoustical Society of America, 142(1), EL7-EL12.
https://doi.org/10.1121/1.4990398
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Jeanne Clarke, Deniz Kazanoğlu, Deniz Başkent, and Etienne Gaudrain
Citation: The Journal of the Acoustical Society of America 142, EL7 (2017); doi: 10.1121/1.4990398 View online: http://dx.doi.org/10.1121/1.4990398
View Table of Contents: http://asa.scitation.org/toc/jas/142/1 Published by the Acoustical Society of America
Articles you may be interested in
Extraction of scattering echo time by surf noise background subtracted autocorrelation The Journal of the Acoustical Society of America 142, EL1 (2017); 10.1121/1.4989996 Intensimetric detection of distortion product otoacoustic emissions with ear canal calibration The Journal of the Acoustical Society of America 142, EL13 (2017); 10.1121/1.4990011 Beyond lexical meaning: The effect of emotional prosody on spoken word recognition The Journal of the Acoustical Society of America 142, EL49 (2017); 10.1121/1.4991328 Comparing malleability of phonetic category between [i] and [u]
The Journal of the Acoustical Society of America 142, EL42 (2017); 10.1121/1.4986422
Modeling speech localization, talker identification, and word recognition in a multi-talker setting The Journal of the Acoustical Society of America 142, 35 (2017); 10.1121/1.4990375
Matched guise effects can be robust to speech style
Effect of F0 contours on top-down repair
of interrupted speech
JeanneClarke,a)DenizKazanoglu,b)DenizBas¸kent,c)and EtienneGaudraind)
Department of Otorhinolaryngology/Head and Neck Surgery, University of Groningen, University Medical Center Groningen, P.O. Box 30.001, BB21, 9700 RB Groningen,
The Netherlands
j.n.clarke@umcg.nl, dkazanoglu@anadolu.edu.tr, d.baskent@umcg.nl, etienne.gaudrain@cnrs.fr
Abstract: Top-down repair of interrupted speech can be influenced by bottom-up acoustic cues such as voice pitch (F0). This study aims to inves-tigate the role of the dynamic information of pitch, i.e., F0 contours, in top-down repair of speech. Intelligibility of sentences interrupted with silence or noise was measured in five F0 contour conditions (inverted, flat, original, exaggerated with a factor of 1.5 and 1.75). The main hypothesis was that manipulating F0 contours would impair linking successive seg-ments of interrupted speech and thus negatively affect top-down repair. Intelligibility of interrupted speech was impaired only by misleading dynamic information (inverted F0 contours). The top-down repair of interrupted speech was not affected by any F0 contours manipulation.
VC2017 Acoustical Society of America [DDO]
Date Received: November 14, 2016 Date Accepted: June 14, 2017
1. Introduction
The brain is able to reconstruct speech partially inaudible, for example, due to back-ground noise. Top-down repair of speech is affected by both linguistic knowledge, expectations, and context (e.g., Bashford et al., 1992), as well as the availability of acoustic bottom-up cues (Bhargava et al., 2014; Clarke et al., 2016). Voice pitch, the perceptual correlate of the fundamental frequency (F0), has been identified as an impor-tant bottom-up cue that helps speech in noise perception. It is a strong across-frequency grouping cue, as (i) pitch information brings coherence to speech sounds by fusing together different parts of the spectrum (e.g., formants), which helps phoneme identifi-cation, and (ii) average F0 and/or dynamic patterns (F0 contours) help to segregate dif-ferent sound sources, which is useful to attend to and better understand the target speech in the presence of maskers. The present study investigates whether voice pitch is also used for sequentially linking successive speech segments across interruptions.
Specifically, the goal of this study was to investigate the effect of the magni-tude and the direction of F0 contours on intelligibility of interrupted speech (with silence and noise) and on phonemic restoration, quantifying top-down repair of speech. The modifications of the original F0 contours (original F0) consisted of (i) inverting the F0 contours within the same magnitude to misrepresent the direction of the F0 dynamic information (inverted F0), (ii) compressing the magnitude of F0 around its median value, thus removing the dynamic information of the F0 contours (flat F0), (iii) expanding the magnitude of F0 by exaggerating the F0 contours with a factor 1.5 (exaggerated 1.5), and (iv) with a factor 1.75 (exaggerated 1.75). These F0 contour manipulations were specifically chosen as they consistently showed an effect on speech perception in background noise (Meister et al., 2011; Miller et al., 2010) or with single-talker interferer (Binns and Culling, 2007).
1.1 Effects of F0 contour manipulations on speech perception and top-down restoration First, the F0 contour manipulations can affect intelligibility of interrupted speech at the speech segment level. Phoneme and coarticulation identification depend on
a)
Also at: University of Groningen, Graduate School of Medical Sciences, Research School of Behavioral and Cognitive Neurosciences, Groningen, The Netherlands.
b)Present address: Anadolu University, Faculty of Health Science, Department of Language and Speech Therapy, Eskis¸ehir, Turkey.
c)
Author to whom correspondance should be addressed. Also at: University of Groningen, Graduate School of Medical Sciences, Research School of Behavioral and Cognitive Neurosciences, Groningen, The Netherlands. d)Also at: Lyon Neuroscience Research Center, Auditory Cognition and Psychoacoustics Team, CNRS UMR
5292, INSERM U1028, University Lyon 1, Lyon, France.
J. Acoust. Soc. Am. 142 (1), July 2017 VC2017 Acoustical Society of America EL7 Clarke et al.: JASA Express Letters [http://dx.doi.org/10.1121/1.4990398] Published Online 5 July 2017
previous language experience, from learned covariation patterns of F0 and formants in speech (Peterson and Barney, 1952). Thus, F0 manipulations not coordinated with unmanipulated formants, such as changing the direction of F0 contours (inverted F0 contour) and reducing F0 contour’s magnitude (flat F0 contour) might produce more identification errors, thus impairing the intelligibility of the speech segments. However, with exaggerated F0 contours, phonemes are more contrasted, and provided they still correspond to proper categories, perception of the independent speech segments could be facilitated (such as in infant-directed speech).
Second, F0 seems to contribute to a robust linking of successive segments of interrupted speech. However, it has been shown that the average F0 of a voice does not partake to the phonemic restoration effect (Clarke et al., 2014). Therefore, here we hypothesize that it is, instead, the dynamic fluctuations of F0 (the contour over time) that are used to link successive speech segments, and thus to restore missing ones. The predictable nature of F0 contours supports that the F0 contour direction would guide the listener to successfully link successive segments of interrupted speech, even more so when the interruptions are filled with noise (Dannenbring, 1976). Inverted contours might impair this linking because of the misleading dynamic information it provides, possibly resulting in a reduced restoration effect, whereas flat contours, without dynamic F0 information, would not help nor impair linking successive speech seg-ments. We were not expecting an exaggerated F0 contour to impair linking successive speech segments (in line with F0 alternation of one octave between successive speech segments not hindering interrupted speech intelligibility and phonemic restoration,
Clarke et al., 2014).
In addition, F0 contours, as a primary feature contributing to prosody, also give information on the intonation of an utterance at the sentential level. Some linguis-tic functions that can be associated with F0 contours are segmentation (word bound-aries) and lexical stress (used for segmentation in English and in Dutch), accentuation (focus on important words in a sentence), and types of utterance (statement or ques-tion), as well as lexical meaning in tonal languages (e.g., Wang et al., 2013). In our study, we expected the inverted F0 contour to have a detrimental effect on intelligibil-ity at the sentential level because lexical stress and accentuation would be misleading (attenuation of important information instead of being highlighted). Moreover, we expected no effect of a flat F0 contour on intelligibility at the sentential level, as the lack of dynamic information could be compensated by linguistic context (in line with
Chatterjee et al., 2010). Furthermore, it is possible that an exaggerated F0 contour might strengthen lexical stress and accentuation, with important information even more highlighted (proportionally with the expansion ratio). We can thus expect a posi-tive effect of exaggerated F0 contour on intelligibility at the sentential level.
Overall, taking into account the expectations at the different levels, i.e., indi-vidual speech segments intelligibility, linking of successive speech segments, and intelli-gibility of the whole sentence, we expected to have better to worse performance for the exaggerated contour, the flat contour, and finally for the inverted contours.
2. Methods
Sixteen native Dutch speakers with normal hearing (20 dB hearing level or less pure-tone thresholds at audiometric frequencies of 250–6000 Hz in both ears), aged between 20 and 40 yrs (mean¼ 25.5, standard deviation ¼ 6.2), and with no hearing or speech-related problems (self-reported), participated in this study. The study was approved by the Medisch Etische Toetsingscommissie (Medical Ethical Review Committee) of the University Medical Center Groningen. All participants were informed about the proce-dure and signed a consent form. Participants were paid for their participation.
The speech stimuli of this study were Dutch everyday sentences with “high sentential-context” (i.e., common vocabulary and semantic context), spoken by a male talker (for more details seeVersfeld et al., 2000). The fundamental frequency (F0) con-tours of the voiced segments of the sentences were manipulated offline using TANDEM-STRAIGHT inMATLAB(Kawahara and Morise, 2011). The five F0 contour
conditions, namely inverted F0, flat F0, original F0, exaggerated F0 by a factor of 1.5, and exaggerated F0 by a factor of 1.75 (see Fig. 1(A), were implemented using the same formula described inBinns and Culling (2007), Miller et al. (2010), andMeister et al. (2011). For the inverted F0 condition, the symmetry about the median F0 value was used in the logarithmic scale (modification ratio a¼ 1). For the flat F0 condition, the median F0 replaced the original F0 (a¼ 0). For the exaggerated F0 by a factor 1.5 and 1.75, the F0 contours were expanded in the logarithmic scale by a ratio 1.5 and 1.75. The resynthesized sentences were modulated with a square wave to obtain an
interrupted version, with an interruption rate of 2.2 Hz, and a duty cycle of 50%, resulting in 227-ms speech segments followed by 227-ms interruptions. Silent and speech shaped noise (signal-to-noise ratio of 5 dB) were used for interruption (as done inClarke et al., 2014).
The participants were seated in a sound-attenuated booth during the experi-ment. The apparatus was the same as used inClarke and colleagues (2014,2016). The responses from the participants were recorded for offline scoring of sentence intelligibil-ity by native Dutch student assistants.
Participants came for a single session which included written informed con-sents, the audiometric test, the baseline measurement, familiarization and data collec-tion, the debriefing, and occasional breaks. To measure the baseline, the first five lists of the stimulus corpus, 13 sentences each, were used without interruption. Each list was presented with a different F0 contour condition, and the order of all lists and con-ditions were randomized for each participant. Participants then familiarized with the procedure listening to five conditions selected randomly out of the ten experimental conditions. The familiarization phase was similar to the experiment expect that written and auditory feedback were provided.
The data collection consisted of 10 conditions: 5 F0 contour conditions (inverted, flat, original, and exaggerated with 1.5 and 1.75 ratio) 2 interruption con-ditions (silent and noise). Both the sentence lists and the concon-ditions were presented in random order. At the beginning of each set the participants heard the same introduc-tion sentence processed with the experimental condiintroduc-tion to come, to prepare them for the trial condition. A tone preceded each sentence to alert the participant. After hear-ing the sentence, the participants were asked to repeat what they could understand from the sentence stimuli, and were additionally encouraged to guess as much as possi-ble. Each word in the sentence was scored according to the participants’ correct response. Total rationalized arcsine transformed unit (RAU) scores were computed for each condition.
3. Results
The upper panel of Fig.1(B)displays the intelligibility scores in each of the ten condi-tions (5 F0 contours interruptions). A repeated-measure two-way analysis of vari-ance (ANOVA) was performed on the RAU scores with F0 contours (5 levels) and interruption (2 levels) as the within-subject factors. The effect size is indicated by eta square, g2. A significant effect of F0 contours [F(4,60)¼ 4.20, p ¼ 0.0046, g2¼ 0.063]
indicated that F0 contours had an overall effect on intelligibility of interrupted speech. A significant effect of interruption [F(1,15)¼ 71.63, p < 0.001, g2¼ 0.15] indicated the
Fig. 1. (Color online) (A) F0 contour manipulations on an example sentence “Buiten is het donker en koud” (Outside it is dark and cold). (B) Intelligibility results (top panel) for silent interruptions (empty boxes) and noise interruptions (filled boxes) and phonemic restoration effect (lower panel) for the manipulated F0 contours, from left to right: inverted (a¼ 1), flat (a ¼ 0), original (a ¼ 1), exaggerated with a factor 1.5 (a ¼ 1.5), and 1.75 (a¼ 1.75). The horizontal line indicates the median, the box indicates the 25th and 75th quartiles, and the dashed whiskers indicate the 1.5 interquartile range. The circles indicate the outliers. The dots indicate the mean.
Clarke et al.: JASA Express Letters [http://dx.doi.org/10.1121/1.4990398] Published Online 5 July 2017
presence of phonemic restoration effect. However, there was no interaction between the two factors [F(4,60)¼ 0.16, p ¼ 0.95, g2¼ 0.0036] which indicated that the F0 con-tour manipulations had the same effect on intelligibility with silent and noise interrup-tions, suggesting that the manipulations of the F0 contours did not affect the difference between the two interruptions (noise and silent) that is a measure of phonemic restora-tion. The phonemic restoration effect was computed by subtracting the scores in the silent condition from those in the noise condition for each F0 contour condition. The results from the phonemic restoration for the manipulated F0 contours are displayed in the lower panel of Fig.1.
The overall effect of F0 contours on intelligibility was small, as shown by the small effect size (g2¼ 0.063) that indicated that only 6% of the variance of intelligibility was explained by the F0 contour manipulations. Differences were observed between only some pairs of conditions. A post hoc analysis using pairwise t-tests with False Discovery Rate (FDR) control was conducted to compare performance of F0 contours between each other (averaged on noise and silent conditions, because of the lack of interaction between the two factors). The results showed that intelligibility performance with inverted F0 contour was significantly poorer from that of other contours [t(15)¼ 4.023, pFDR¼ 0.0034; t(15) ¼ 2.84, pFDR¼ 0.026; t(15) ¼ 3.60, pFDR¼ 0.0054,
for original, exaggerated F0 contour with ratio 1.5 and 1.75, respectively] except that of the flat F0 contour [t(15)¼ 2.0075, p ¼ 0.13]. All other comparisons were not signifi-cantly different. Moreover, the baseline scores (of uninterrupted sentences) were at ceil-ing and were not affected by the F0 contour manipulations [F(4,18)¼ 0.481, p ¼ 0.749, g2¼ 0.013], suggesting that our F0 contour manipulations did not impair intelligibility of uninterrupted speech.
There was no effect of F0 contours on the phonemic restoration benefit as indicated by the lack of interaction in the ANOVA on the intelligibility scores. However, as an a priori variable, we tested whether the phonemic restoration scores were significantly different from 0 with a one-sample t-test for each F0 contour condi-tion (indicated by a black star on the lower panel of Fig. 1). A significant phonemic restoration benefit was observed in all F0 contour conditions [t(15)¼ 3.87, p ¼ 0.0015 for inverted; t(15)¼ 2.90, p ¼ 0.011, for flat; t(15) ¼ 2.53, p ¼ 0.023, for original; t(15)¼ 2.88, p ¼ 0.012, for exaggerated F0 contour with ratio 1.75], except one, the exaggerated F0 contour with ratio 1.5 [t(15)¼ 2.022, p ¼ 0.061]. This indicates that F0 contour manipulations did not affect top-down repair mechanisms of interrupted speech.
4. Discussion
We were interested in investigating the effects of the magnitude and direction of F0 contours on intelligibility and top-down repair of interrupted speech. We showed that modifying the magnitude of F0 contours (all conditions except the inverted contours) did not have any effect on interrupted speech intelligibility, contrary to other studies with continuous background interferer (Binns and Culling, 2007; Miller et al., 2010;
Wang et al., 2013). However, as already pointed out for interrupted speech with silence,Chatterjee et al. (2010)did not show reduced intelligibility with flat F0 contour at a faster interruption rate, suggesting again the prevalence of linguistic cues. The results from the two exaggerated F0 contours did not confirm our hypothesis on a bet-ter phonetic categorization improving speech perception with wider F0 variations, but confirmed our hypothesis that wider F0 variations do not weaken linking successive speech segments (in line with Clarke et al., 2014). On the other hand, the direction of the F0 contours seems to be a cue listeners relied on for intelligibility of interrupted speech. Indeed, partially validating our hypothesis, having misleading dynamic infor-mation of F0 (inverted contours) impaired interrupted speech intelligibility. This con-firms that wrong F0 dynamic information can lead to lower intelligibility of interrupted speech, as was found for speech with continuous background interferer (Binns and Culling, 2007; Meister et al., 2011; Miller et al., 2010). An explanation can be that original F0 contour helps to define clause boundaries whereas inverted F0 contour dis-torts those boundaries (Wingfield et al., 1984), impairing sentence intelligibility. Taken all together, these results suggest that the F0 variations’ magnitude may not be used as a linking cue for interrupted speech perception, but that direction of F0 contours may. Moreover, only when all three aspects of interrupted speech perception (individual speech segments intelligibility, linking successive speech segments, and sentential intelli-gibility) were affected by the F0 contour manipulation, i.e., for the inverted F0 contour, did overall intelligibility decrease. This suggests that participants seem to fail
to compensate for the inverted F0 contour manipulation that impairs more aspects of interrupted speech perception than our other F0 contour manipulations.
Nevertheless, even the misleading dynamic information of F0 did not impair the top-down repair of interrupted speech. This result suggests that participants may have compensated for the atypical F0 cues with the linguistic context (as suggested by
Chatterjee et al., 2010). In this study, it seems that top-down repair of speech may rely more on linguistic cues than on F0 cues (Clarke et al., 2014). For F0 contours’ magni-tude (flat and exaggerated F0 contour conditions), the restoration benefit remained, suggesting that sequentially linking successive speech segments was not affected by F0 contours’ magnitude, likely because of the prevalence of the linguistic contextual cues in the present task. However, F0 variations may be necessary for other tasks, such as recognizing emotions, and might thus be more difficult to compensate for in such a case. We can thus speculate that with “low-context” sentences (no semantic context but syntactically correct, for example), the task becomes harder, and F0 contour manipulations might further impair intelligibility and top-down-repair of interrupted speech. Furthermore, when F0 contours were exaggerated, the noise bursts filling the silent interruptions still acted as a masker with wider F0 variations across the noise bursts (in line withClarke et al., 2014). The present study suggests that the F0 contour manipulations, presumably weakening the sequential linking of successive speech seg-ments (for inverted F0 contour), did not affect phonemic restoration benefit. Thus, for interrupted speech, it seems that speech segments with F0 contour manipulations can still clearly be discriminated from the filler noise, a mechanism involved in top-down repair of speech. In contrast, in the case of a speech masker, discriminating competing talkers (target and maskers) relies on F0 contours. Presumably, discriminating two sources of same nature (speech-on-speech) is more affected by F0 contour manipula-tions than discriminating two sources different in nature (speech and noise) as sug-gested by the difference of results observed between the present study and previous studies (Binns and Culling, 2007;Meister et al., 2011;Miller et al., 2010).
Even if interrupted speech intelligibility did not significantly differ between F0 contour manipulations (except for the inverted contours), it is still possible that our participants did require more effort to perform the task with the atypical F0 contours. Moreover, speech redundancy is present at different layers of speech processing, and other cues, instead of F0 contours, might be used for prosody processing, such as dura-tion and intensity. Depending on the task difficulty (affected by the amount of infor-mation in the speech stimuli), listeners may rely differently on prosodic inforinfor-mation. For example, the redundancy added to speech from normal prosody may be relevant when the task becomes harder by reducing the processing time (i.e., increasing the cog-nitive load), especially using duration cues in prosody (a deficit of flat F0 over normal F0 being observed at normal speech rate and not for time-compressed speech in
Wingfield et al., 1984). This is in line with the fact that other cues, instead of F0 con-tours, might be used for prosody processing. Indeed, intensity and duration are good indicators of prosodic information as they co-vary with F0 contours and provide redundant information for prosody processing. As a result, even when F0 contours are manipulated, intensity and duration can be used for stress perception, segmentation, and intonation recognition (e.g., Peng et al., 2012). In the present study, the inverted F0 contour, which provided misleading and distorted cues, did not complement speech redundancy, which might explain the lower performance in speech intelligibility for this condition.
To summarize, the present study shows a relatively small effect of F0 contour manipulations on intelligibility of interrupted speech and no effect on phonemic resto-ration. Confirming the findings of Clarke et al. (2014), these results indicate that top-down repair of speech could be robust to atypical voice cues, suggesting that listeners may partly compensate for the degraded voice cues. It is possible that linguistic infor-mation, such as the sentential context, which plays an important role in the restoration mechanisms, helped overcome the negative effects of manipulated F0 contours. Another possibility is that participants relied on other co-varying prosodic cues, such as intensity and duration. Presumably, a combination of both mechanisms may occur to achieve the best possible performance.
Acknowledgments
The authors would like to thank Floor Burgerhof and Wilke Bosma for transcribing participant responses, as well as the participants. This study was supported by a VIDI grant from the Netherlands Organization for Scientific Research, NWO, and Netherlands Organization for Health Research and Development, ZonMw (Grant No. 016.096.397).
Clarke et al.: JASA Express Letters [http://dx.doi.org/10.1121/1.4990398] Published Online 5 July 2017
Further support came from a Rosalind Franklin Fellowship from the University of Groningen, University Medical Center Groningen, and funds from Heinsius Houbolt Foundation. The study is part of the “Healthy Aging and Communication” research program of the Otorhinolaryngology Department of University Medical Center Groningen and was also conducted in the framework of the LabEx CeLyA (“Centre Lyonnais d’Acoustique,” ANR-10-LABX-0060/ANR-11-IDEX-0007) operated by the French National Research Agency.
References and links
Bashford, J. A., Riener, K. R., and Warren, R. M. (1992). “Increasing the intelligibility of speech through multiple phonemic restorations,”Percept. Psychophys.51, 211–217.
Bhargava, P., Gaudrain, E., and Bas¸kent, D. (2014). “Top-down restoration of speech in cochlear-implant users,”Hear. Res.309, 113–123.
Binns, C., and Culling, J. F. (2007). “The role of fundamental frequency contours in the perception of speech against interfering speech,”J. Acoust. Soc. Am.122, 1765–1776.
Chatterjee, M., Peredo, F., Nelson, D., and Bas¸kent, D. (2010). “Recognition of interrupted sentences under conditions of spectral degradation,”J. Acoust. Soc. Am.127, EL37–EL41.
Clarke, J., Gaudrain, E., Chatterjee, M., and Bas¸kent, D. (2014). “T’ain’t the way you say it, it’s what you say—Perceptual continuity of voice and top-down restoration of speech,”Hear. Res.315, 80–87. Clarke, J. N., Bas¸kent, D., and Gaudrain, E. (2016). “Pitch and spectral resolution: A systematic
compari-son of bottom-up cues for top-down repair of degraded speech,”J. Acoust. Soc. Am.139, 395–405. Dannenbring, G. L. (1976). “Perceived auditory continuity with alternately rising and falling frequency
transitions,”Can. J. Psychol.30, 99–114.
Kawahara, H., and Morise, M. (2011). “Technical foundations of TANDEM-STRAIGHT, a speech anal-ysis, modification and synthesis framework,”Sadhana36, 713–722.
Meister, H., Landwehr, M., Pyschny, V., Grugel, L., and Walger, M. (2011). “Use of intonation contours for speech recognition in noise by cochlear implant recipients,”J. Acoust. Soc. Am.129, EL204–EL209. Miller, S. E., Schlauch, R. S., and Watson, P. J. (2010). “The effects of fundamental frequency contour
manipulations on speech intelligibility in background noise,”J. Acoust. Soc. Am.128, 435–443. Peng, S.-C., Chatterjee, M., and Lu, N. (2012). “Acoustic cue integration in speech intonation recognition
with cochlear implants,”Trends Amplif.16, 67–82.
Peterson, G. E., and Barney, H. L. (1952). “Control methods used in a study of the vowels,”J. Acoust. Soc. Am.24, 175–184.
Versfeld, N. J., Daalder, L., Festen, J. M., and Houtgast, T. (2000). “Method for the selection of sentence materials for efficient measurement of the speech reception threshold,” J. Acoust. Soc. Am. 107, 1671–1684.
Wang, J., Shu, H., Zhang, L., Liu, Z., and Zhang, Y. (2013). “The roles of fundamental frequency con-tours and sentence context in Mandarin Chinese speech intelligibility,” J. Acoust. Soc. Am. 134, EL91–EL97.
Wingfield, A., Lombardi, L., and Sokol, S. (1984). “Prosodic features and the intelligibility of accelerated speech: Syntactic versus periodic segmentation,”J. Speech Lang. Hear. Res.27, 128–134.
